Smart Engines introduced a new generation of identity documents recognition systems for scanning passports, ID cards, bank cards, barcodes and other documents, checking their authenticity and performing biometric verification. Nowadays, the company has become the universal software supplier for precise instant and secure data extraction, document liveness detection, computational document forensics and face matching for users’ verification. The new products, Smart ID Engine, Smart Code Engine and Smart Document Engine, have been designed to meet Green AI principles and are aimed to provide a higher level of privacy protection and security for users and businesses against document fraud.
Smart ID Engine is a complete renovation of the company’s flagship technology Smart IDReader and it enables to extract data from more than 1483 ID cards, passports, driver’s licenses and other documents of 210 jurisdictions worldwide with support for 99 languages, including Arabic, Persian, Urdu, Chinese, Japanese, Korean, and many more. Smart Code Engine combines solutions for optical reading (scanning) of credit and debit cards, barcodes and machine-readable zone (MRZ). Smart Document Engine is a system for automatic classification, recognition and extraction of the required data from structured, semi-structured and unstructured documents: tax certificates, bills, accounting and audit documentation, payment orders, and more. The new software solutions can be integrated into mobile, desktop and server applications and work autonomously and independently, ensuring the security of personal and sensitive data processing.
“Given that personal data leaks are becoming commonplace, user protection against personal data abuse and data leakage images are becoming the main focus everywhere. Developing new products, our efforts were aimed at improving the recognition quality and speed, as well as preventing various document fraud and data theft scenarios,” says Vladimir Arlazarov, CEO of Smart Engines, PhD.
The presented solutions were updated with the proprietary AI technology GreenOCR ®, in which a new 8-bit inference model for deep neural network inference was implemented. The technology was created under the Green AI initiative and the sustainable development program led by the UN. The 2-nd generation of GreenOCR ® allows you to significantly reduce power consumption during recognition, as well as process documents faster. The recognition performance on powerful modern smartphones has increased by 10%, and user experience on low-end and medium-end devices has improved by 30%. Thanks to the new Deep Visual Memory technology, modular architecture, additional algorithmic and technical optimizations, we managed to reduce the distribution size increasing the number of new documents and their templates. The distribution is reduced by 25% for US and European ID cards, 15% for passports worldwide, 35% for bank cards and 40% for barcodes. These innovations reduce the carbon footprint and extend the life of mobile devices on a single battery charge.
The main feature of the new products is the support for the Arabic script recognition and languages of the Indo-Iranian group (Persian (Farsi), Urdu, Tamil, Sinhalese languages), which significantly broadens the marketing target in the Middle East, South-East Asia and Africa. It means 21 jurisdictions with a total population of 500 million people who do not use a Latin-based writing system in national documents. The Arabic language recognition is implemented for 73 types of documents, including passports, ID cards and driver’s licenses of the following countries: Algeria, Bahrain, the State of Palestine, Egypt, Jordan, Iraq, Iran, Yemen, Qatar, the Comoros, Kuwait, Lebanon, Libya, Mauritania, Morocco, the United Arab Emirates, Oman, Saudi Arabia, Syria, Sudan, Tunisia.
At present, Smart ID Engine enables quick document recognition in the video stream, photos and images from scanners. The recognition time for the US driver’s license is about 250 ms per 1 frame and the German ID card is about 150 ms. The solution always works independently on the end device (smartphone, smart camera, terminal, computer, server), does not require a network connection and does not transmit client data for processing to third-party services and/or third parties for manual data entry. Smart ID Engine helps companies to comply with the security standards for personal data processing (GDPR, CCPA, and others) and satisfy the strict demands of regulators in terms of KYC/AML.
Automatic document type identification and SDK compact configuration (or a set of libraries) are immediately available for users and integrators in Smart ID Engine. These features help to extract data promptly in an automatic mode without a specified document type for passports of 211 jurisdictions, ID cards and driver licenses of all EU countries and Latin America, driver licenses of all US States, all ID documents of the EU and CIS countries.
During recognition in the video stream, Smart ID Engine is able to select the best frame of a document image to save it in the company’s accounting system and the least damaged (the clearest and glare-free) photo of the face in the document, which is extracted as a separate item. Biometric user’s verification implemented in the new products is based on the company’s proprietary face recognition technology developed by the Smart Engines research team, which allows you to perform face matching, compare a face photo in a document with the owner’s face in another photo, including a selfie taken by the user.
Smart ID Engine integrated methods for document liveness detection automatically reveal attempts to use forged documents, photocopies, as well as ID image shot from the screen of a computer monitor, tablet or smartphone. AI algorithms detect holographic security elements, special visual security elements (monograms, ornaments, etc.), check the document page geometry and control the document state in the video stream and in individual frames. By a principle of work, liveness detection belongs to the “passive” class, it means it does not require the user to perform special activities. To recognize and verify a document, you need only show it to the camera device for a few seconds or just take a photo.
“The development of online industries and remote services have forced the world to take a new look at document processing with personal and sensitive data using mobile applications and web services. Liveness document verification based on the principle of liveness biometric verification for a person’s face is a vital improvement of anti-fraud solutions, which perfectly complements existing technologies for document fraud prevention and confirming their belonging to the user. It enhances considerably the protection of end users from possible misuse of their documents,” said Andrew Sever, CEO and co-founder of Identity verification platform Sumsub.
To verify the validity of the document, the date of birth, date of issue and document validity period are compared with the recognition date. Updated authenticity verification includes the identification of fraud signs on the document forms and documents of foreign countries.
Smart Code Engine provides automatic data reading of bank cards, machine-readable zone (MRZ) and barcodes in different image capture conditions. The system is resistant to lighting condition changes, shooting angle, skewness, angular rotation, excessive and insufficient lighting. The system works in real time and is able to recognize bank cards and other objects on modern devices at a speed of 25 frames per second.
In payment services, Smart Code Engine allows you to scan and recognize debit and credit cards of MIR, VISA, MasterCard, American Express, JCB, Maestro, UnionPay and Diners Club payment systems issued by various countries worldwide, providing extraction of not only the card number (12-19 digits), but also the expiry date and owner’s name. Bank card recognition of all types is supported: embossed, indent and flat printed, with horizontal or off-standard vertical ID data alignment and cards with lettering on averse or reverse side of the card. In the new product, now users can recognize bank cards with the IBAN number, which are widely used in the European Union.
The Smart Engines solution provides automatic search and data recognition from machine-readable document zones (MRZ) made in accordance with the established international standards ISO / ICAO (IEC 7501-1/ICAO Document 9303 ISO) and local (Russia, France, Switzerland, Bulgaria, Ecuador) standards on passports, ID cards, visas, driver licenses and other documents. Liveness detection tools and data consistency are available for Smart Code Engine users recognizing MRZ documents.
Smart Code Engine provides high-quality recognition for one-and two-dimensional barcodes from various invoices and receipts, including utility and government bills, tax documents and invoices, as well as tickets, receipts, invoices, price tags, posters and ads. The new product has improved AZTEC code recognition by 1.5 times due to the implementation of a new technology for barcode model detection and PDF417 code, including their data compaction modes. The QR code reader developed by Smart Engines is now able to detect inverted codes and is also more resistant to their orientation. Today, the product supports recognition for QR Code, AZTEC, PDF 417, Data Matrix, CODABAR, CODE_39, CODE_93, CODE_128, EAN_8, EAN_13, ITF, UPC_A, UPC_E.
“Smart Engines recognition technologies provide business with a unique opportunity to effectively implement new approaches without high requirements to image capture and hardware for performing AI algorithms. Users can take photos or show a document, bank card or barcode to the camera as they like, while all calculations do not require high-performance work systems and take seconds even on budget phones,” said Vyacheslav Tsyganov, VP and CIO Tinkoff Bank.
Smart Document Engine solves problems of automatic data extraction from standard document forms, strict accountability documents, accounting, financial, tax, legal, notarial and other documents used in document workflow, various tests and KYC questionnaires, on scans and photos. The system allows you to automatically process single-and multi-page documents with fixed and flexible layouts, documents with floating blocks and details, unstructured text documents and blocks, tables, labels or even individual text lines and markings.
The software product allows you not only to quickly recognize data from KYC questionnaires, forms and documents, but also to verify them for compliance with formalities. Smart Document Engine can check whether there is a signature, stamp or logo, whether they are the right color, whether they are in the right document place, and check that the inscriptions that should be made by hand are really handwritten. In addition, it is possible to verify the integrity and immutability of the form, document or its part during processing. Due to second-generation GreenOCR ® technology, the processing time of A4 document page on AMD Ryzen 7 3700X is about 2 seconds.
To enable easy integration, .NET wrappers for Linux systems, new improved interfaces for C++ and Java, and React and Flutter support for mobile SDKs are available for developers now. The whole product range is optimized for X86, ARM, MIPS and SPARC, architectures and is compatible with iOS, Android, Linux, Windows, MacOS, Aurora, Solaris and Elbrus operating systems. The modular software architecture minimizes the size of the software library, because it allows configuring the SDK for a specific customer and including only the necessary functionality.
Smart Engines software solutions successfully solve the digital transformation problems in organizations of various industries worldwide. Technologies are used by Tinkoff group, Citibank, ROSBANK (Societe Generale Group), Home Credit Bank, Beeline, MTS, BioCollections Worldwide, BlaBlaCar, Blockpass IDN Ltd, Dukascopy Bank SA, Tessi, VerifyMyAge, etc.
“Taking care of the environment is one of the Smart Engines key priorities. Our research in the field of low-bit and bipolar neural network architectures is devoted to the environmental protection, the use of which allows us to minimize the carbon footprint of our AI models at the training and execution stages. It is important to emphasize that the company’s scientists and engineers managed to improve the recognition quality, add new functionality, improve the user experience and significantly reduce power consumption in the new solutions,” said Vladimir Arlazarov, CEO of Smart Engines, PhD.